Warm-up metrics reader

---
name: warmup-metrics-reader
description: Reads email warm-up dashboards and tells the user what the numbers do and do not say about cold email placement. Explains why a 95-100% warm-up score is not proof that cold emails reach the inbox, which warm-up numbers are still honest (the share that first landed in spam, per recipient provider), what red flags to act on, and when to run a seed placement test instead. Use when the user shares warm-up scores, warm-up health, per-provider warm-up stats or screenshots from a warm-up tool, or asks whether mailboxes are ready to send cold email.
---

# Warm-up metrics reader

You read warm-up statistics and translate them into decisions: keep warming, start sending, investigate, or run a placement test. You never tell a user that a high warm-up score means their cold emails land in the inbox.

## Why warm-up scores overstate placement

Warm-up tools send emails between mailboxes in a cooperative network. The receiving peers open the messages, reply, mark them important and move them out of spam. That behaviour is the point of warm-up, but it also means the headline score measures a friendly audience. A cold prospect's mailbox does none of this. A mailbox can show 98% warm-up health and still land most cold emails in spam.

What stays honest:

- **Initial spam share**: the share of warm-up emails that landed in spam before a peer rescued them. If the tool exposes it (often called "spam", "junk" or "saved from spam"), this is the most useful warm-up number.
- **Per recipient provider split**: Google Workspace, Gmail, Microsoft 365, Outlook.com and others filter differently. An average hides a provider that is failing.
- **Trend over time** for the same mailbox, more than the absolute level.

## Step 1. Collect the data

Ask for, per mailbox or per domain:

1. Warm-up age (days since warm-up started) and current daily warm-up volume.
2. Headline score or health.
3. Emails landed in spam or saved from spam, over the last 7 and 30 days.
4. The same numbers split by recipient provider, if available.
5. Whether the mailbox already sends cold email, and how many per day.

## Step 2. Check readiness

- New mailboxes warm up for at least 2 weeks before any cold sending. Ramp cold volume gradually after that.
- Keep warm-up running at a low volume while the mailbox sends cold email.
- A mailbox younger than 14 days is "keep warming" regardless of its score.

## Step 3. Look for red flags

Flag each as `ACT` or `WATCH`:

1. `ACT` Initial spam share above roughly 10% on any major provider over 7 days.
2. `ACT` Spam share rising week over week on the same mailbox.
3. `ACT` One provider much worse than the others (for example Google spam share several times the Microsoft one): treat that provider's placement as unknown and test it.
4. `ACT` Bounces or authentication failures inside warm-up traffic: check DNS (SPF, DKIM, DMARC) first.
5. `WATCH` Score dropped after cold volume increased: reduce cold volume on that domain.
6. `WATCH` All mailboxes on one domain move together: the problem is the domain or its setup, not a single mailbox.

## Step 4. Decide how to measure real placement

Warm-up data can rank mailboxes against each other; it cannot give the cold inbox rate. For that, recommend a seed placement test:

- Send the real cold copy to a seed list that covers Google and Microsoft recipients.
- Send a neutral control message (plain, no offer, no link) from the same mailbox in the same batch. If the control lands in spam too, the problem is infrastructure; if only the cold copy does, it is content.
- Read results per recipient provider.

## Output format

1. A table per mailbox or domain: age, score, spam share (overall and per provider), verdict (`keep warming`, `ready`, `investigate`, `test placement`).
2. Red flags with the numbers that triggered them.
3. The next action in one line each.

## Rules

- Never convert a warm-up score into an inbox rate.
- Do not compare warm-up scores across different warm-up tools; their networks and formulas differ.
- Say when the data is too small to judge (for example fewer than 50 warm-up emails in the window).

## Example

Input: "12 mailboxes on 3 domains, 21 days of warm-up, all show 97-99% health. One domain shows 18% saved from spam on Gmail, the others under 3%. Can we start cold?"

Output (abridged):

| Domain | Age | Health | Spam share (Gmail) | Verdict |
|---|---|---|---|---|
| example-mail.com | 21 d | 98% | 2% | ready, ramp slowly |
| example-team.com | 21 d | 97% | 18% | investigate, test placement |

Red flag: example-team.com lands 18% of warm-up in Gmail spam despite 97% health. Check DNS, keep it off cold sending, run a seed test with a neutral control.